Sleep Electroencephalogram in Seasonal Disorder and in Control Women: Effects Light Treatment and Sleep Deprivation

Daniel P. Brunner, Kurt Kr~iuchi, Derk-Jan Dijk, Georg Leonhardt,

Hans-Joachim Haug, and Anna Wirz-Justice

Affective of Midday

The role of sleep regulation in Seasonal Affective Disorder (SAD) was studied in 11 female SAD patients and eight controls in winter before and after light treatment (LT, 6000 lux, 10-14h, 5 days). The sleep electroencephalogram (EEG) was recorded at baseline and after the total sleep deprivation (TSD) of a 40-h constant routine. The well-known effects of TSD on sleep parameters and on EEG power spectra were replicated, indicating normal homeostatic sleep regulation in SAD. Sleep improved after LT in both groups. Since the condition following LT was the second session, these improvements may be an order effect and~or an effect of LT itself After LT, sleep EEG spectra of SAD patients, but not of controls, showed modifications resembling those of recovery sleep. Since only SAD patients curtailed their sleep while remitting during the LT period, these EEG modifications can be explained by normal sleep regulation alone. We conclude that the robust antidepressant effect of LT in SAD is unlikely to be mediated by changes in sleep, and that sleep regulatory mechanisms are not a crucial factor in the pathogenesis of winter depression.

Key Words: Seasonal affective disorder, sleep EEG, light treatment, sleep deprivation, depression, spectral analysis, sleep homeostasis

BIOL PSYCHIATRY 1996;40:485--496

Introduction The winter recurrence of depressive episodes in patients suffering from Seasonal Affective Disorder (SAD) sug- gests that a biological mechanism sensitive to seasonal changes is linked with this particular form of depression. Patients suffering from the winter type of SAD frequently report hypersonmia, fatigue, loss of energy, carbohydrate

From the Psychiatric University Clinic (DPB, KK, GL, H-JH, AW-J), University of Basel, and the Institute of Pharmacology (D-JD), University of Ztirich, Switzerland.

Address reprint requests to Prof. Anna Wirz-Justice, Ph.D., Psychiatric University Clinic, Wilhelm Klein-Strasse 27, CH-4025 Basel, Switzerland.

Received March 6, 1995; revised September 22, 1995.

craving, appetite and weight gain (Rosenthal et al 1984). The temporal link of the disorder to the winter months and the dependence of its prevalence on geographical latitude (Rosen et al 1990; Wirz-Justice 1994a) suggest that light exposure could be a key factor in the pathogenesis of SAD. Support for this assumption is also lent by a metaanalysis showing that 60% of the patients respond clinically to bright light treatment (Terman et al 1989). After a decade of clinical research, bright light is consid- ered the treatment of choice for winter SAD (Society of Light Treatment and Biological Rhythms 1990; Terman and Terman 1991), however, there is still controversy as to mechanisms of action (Terman 1993; Wirz-Justice 1994a)

© 1996 Society of Biological Psychiatry 0006-3223/96/$15.00 SSDI 0006-3223(95)00656-7

486 BIOL PSYCIMATRY D.P. Brunner et al 1996;40:485-496

including the extent of placebo effects (Eastman et al 1993). Since depressive symptoms in SAD patients can be reversed within a few days of light treatment (LT), this group of depressed patients is ideal to investigate interre- lationships between mood and physiology.

Hypotheses about the pathophysiology of SAD have focused on abnormalities in circadian pacemaker function, such as a phase delay (Lewy et al 1987), a diminished amplitude of circadian rhythms during winter (Czeisler et al 1987, 1989), or abnormal photoperiodic response (Rosenthal et al 1984). Accordingly, the therapeutic ef- fects of bright light have been attributed to its ability to normalize aberrant circadian rhythms. Since the circadian pacemaker is a major determinant of sleep propensity, sleep timing, and sleep structure (Czeisler et al 1980; Zulley et al 1981), light-induced modifications of the pacemaker may also have repercussions on sleep. The changes in sleep may then affect depressed mood, as a close link between the regulation of sleep and depressive mood is well known. On one hand, a decline of sleep quality and characteristic changes in rapid eye movement (REM) sleep and nonREM sleep occur during major depression (Gillin et al 1979; Reynolds and Kupfer 1987; Benca et al 1992). On the other hand, rapid improvement of depressive symptoms with sleep deprivation and the rapid relapse after recovery sleep lend strong evidence for common regulatory mechanisms of sleep and mood (Kuhs and Toelle 1991; Wirz-Justice 1994b). It has been hypoth- esized that an alteration of the homeostatic aspect of sleep regulation might account for the simultaneous changes in sleep and depressive mood (Borb61y and Wirz-Justice 1982). According to this model, the antidepressant re- sponse to LT is expected to result from the restoration of a presumably deficient homeostatic regulation of nonREM sleep.

A number of studies have investigated sleep structure and sleep electroencephalogram (EEG) in depressed SAD patients before and after LT (Rosenthal et al 1989; Endo 1993; Partonen et al 1993; Anderson et al 1994; Kohsaka et al 1994; Palchikov et al 1996). None found the typical pattem of EEG changes characterizing sleep in major depression, nor did they report consistent or large effects of bright LT on sleep structure. These studies did not address the question of whether depressed SAD patients have an abnormal homeostatic regulation of sleep. Fur- thermore, only one of the previous investigations used a control group to test whether sleep changes after LT are specific for SAD patients.

In the present study, we determined whether changes in circadian and/or homeostatic mechanisms of sleep regula- tion occur after LT and whether they are similar for SAD patients and healthy subjects. To that aim, visual and spectral analysis of sleep EEG recordings contiguous to a

40 h constant routine protocol were performed prior to and following midday LT.

Methods

Subjects and Study Design From a population of more than 1000 women, who responded to advertisements in the lay press and returned a seasonal screening questionnaire, 11 women diagnosed with winter SAD (Rosenthal et al 1984) and eight control women experiencing no seasonality and reporting no history of depression, participated in the study as paid volunteers. The SAD patient collective was not selected for hypersomnia, and indeed, only four of the 11 patients indicated that they slept at least one hour longer during winter depression. Nevertheless, all but one SAD patient described a moderate to marked increase in subjective sleep need. All subjects signed a written informed consent after all experimental procedures had been explained. SAD patients were between 27 and 66 years old (mean _ SD: 46.5 ___ 12.9) and the controls between 24 and 66 years (50.4 ___ 12.5). Body mass index and demographic variables were very similar between the two groups. Five SAD patients and three of the control subjects were premenopausal and entered the protocol between the first and fourth day of the menstrual cycle in order to complete the entire experiment within the follicular phase. All subjects were non-smokers, free of medication including contraceptives, and underwent a medical examination prior to the study. A three-month period free of shift work, time zone travels, and antidepressant medication was required. They were requested to refrain from alcohol and caffeine for the ten days of the experiment and to reduce the consumption of such beverages in the week prior to the study.

The experiments were carded out from October to March, with the majority of the women participating between November and February. The experiment lasted for 10 days and included two pairs of nightly sleep recordings separated by 5 days of LT at home. The subjects were instructed to sit in front of a portable light box from 10:00h to 14:00h each day. They were exposed to 6000 lux at 60 cm from the light source which consisted of 10 fluorescent light tubes with a total illumination surface of 55 × 40 cm.

Immediately prior to and following the 5 days of LT all subjects spent 21/2 days in the sleep laboratory in order to undergo two identical Constant Routine (CR) protocols of 40 hrs. Both CR protocols started after a night of baseline sleep and ended with the beginning of recovery sleep. The design of the study is visualized in Figure 1. The subjects

Sleep EEG in SAD: Effects of Light and TSD BIOL PSYCHIATRY 487 1996;40:485-496

40 h Constant Routine 40 h Constant Routine

5 DAYS of LT 6000 lux; 10:00-14:00h

BL REC BL REC

Figure l. Study design consisting of two identical constant routine protocols starting with a baseline (BL) sleep episode and ending with recovery sleep (REC). Five days of bright light treatment (LT) separate the two experimental sessions.

informed us of their habitual sleep times and for the entire study these were fixed for each individual and kept constant for the four sleep episodes in the laboratory. They were requested to refrain from napping prior to both baseline nights. The subjects kept a detailed sleep diary one week before, during, and after the experiment to document compliance to the above instructions. Acto- grams collected by wrist-worn activity monitors further allowed inspection of their sleep patterns. All women spent the entire time on the two CR protocols sitting in bed being kept awake and busy by a technician who enforced adherence to the stringently timed procedures of the CR protocol. The procedure of the CR was adapted from Czeisler et al (1989). Sustained wakefulness in a constant propped-up body position (45 ° angle) and under controlled room conditions (temperature: 22°C, humidity: 67%, light: < 80 lux) represents a very standardized total sleep deprivation. Subjects reported to the laboratory in the evening before each baseline night and completed ques- tionnaires and performance tests in bed in order to famil- iarize themselves with the procedures of the CR-protocol which started the next morning. Apart from the instruction about the timing of LT, no restrictions about daytime activities were imposed on the subjects. A few days prior to the first experimental sleep episode an adaptation night was scheduled for adjustment to the new sleeping envi- ronment and the recording equipment.

Depression was assessed in a number of ways, in order to document any changes during the sleep deprivation of the CR and after light treatment (Hang et al 1995; Wirz-Justice 1994b). SAD patients were required to have a 21-item Hamilton Depression Rating (HAM-D, Hamil- ton 1967) of greater or equal 13 at interview, and to have confirmed that the characteristic onset of seasonal symp- toms had begun. Atypical symptoms were assessed by an addendum scale specifically developed for winter depres- sion (Rosenthal et al 1984). In three SAD patients the HAM-D rating fell below 13 points during the week prior to the baseline sleep recording. Due to the enormous personal engagement required to participate in the de- manding protocol, a selection biased towards mildly de- pressed patients could not be avoided. Both SAD patients and controls completed 6-hourly self-ratings of depression (DS/DS'; von Zerssen and Koeller 1976) as well as

half-hourly mood ratings on a visual analogue scale (VAS; Aitken 1969) throughout each CR protocol. Quite small changes in the latter self-rating have been shown to be a reliable estimate of a clinically observed improvement (Haug and F~ihndrich 1986).

The study was not conceived as a clinical trial of light treatment, but as a controlled trial of the effects of a high dose of bright light on multiple physiological parameters. Thus, there was no blind in the design, nor a control for placebo effects.

Sleep Recordings and Data Analysis

Two EEG signals (C3-A2 and C4-A1), two EOG signals, and one EMG and EKG signal were continuously recorded on paper (10mm/sec; Nihon Kohden). All records were scored for 20-s epochs according to the criteria of Re- chtschaffen and Kales (1968). The time constant of the EEG channels was set to 1 sec and the low pass filter to 35Hz (12dB/octave). The amplifier output of the EEG signal was recorded on tape and conveyed on-line or off-line to a signal processor card installed on a personal computer. Analog to digital conversion with a sampling rate of 128 Hz and a 12 bit resolution (ca. 0.15pN/LSB) and the calculation of power spectra were performed by the signal processor card. Power spectra were calculated for consecutive 4-s epochs and 0.25-Hz band widths by a fast fourier transformation routine using a Kaiser-Bessel spectral window (alpha = 3.5). The values of adjacent 0.25-Hz bands were collapsed into 0.5-Hz bins between 0.25 and 5.0 Hz, and into 1.0-Hz bins between 5.25 and 25.0 Hz. A time-mark generated in connection with the on-line signal processing was written on the polygraph paper at 20-s intervals to enable synchronization of the sleep stage scoring with the spectral data. In those in- stances when the EEG spectra were calculated off-line from tape, time alignment with the sleep staging was achieved by the following procedure. Values of mean power density in the 22.25-25.0 Hz band were printed out for consecutive 4-s epochs. On this print-out, surges of high frequency activity in the EEG during movement arousals were easily identified. The paper record was then manually numbered into 20-s epochs by keeping the movement artifacts on the EEG paper lined up (in time)

488 BIOL PSYCHIATRY D.P. Brunner et al 1996;40:485-496

Table1 .

SAD-patients (N=I 1) Control women (N=8)

before LT after LT t-Test

Hamilton (HAM-D)

Atypical Items

DS/DS' before CR

(at 21:00h)

DS/DS' during CR

(mean of 9 ratings)

VAS: 100mm=very good

0ram=depressed mood

before LT

13.9+5.0

7.5±3.2

16.0*-6.9

15.3±6.2

38.84-6.7

after LT

3.14-1.6

1.6±2.1

8.04-3.9

9.2+3.4

50.04-11.2

t-Test

0.0001

0.0002

0.0003 3.34-2.2

0.003 3.5±2.4

0.02 71.2±11.8

3.04-1.7

2.64-2.2

69.3+14.8

t-l.S.

n . s .

n . s .

Hamilton Depression Ratings (HAM-D), ratings of atypical depressive symptoms, self ratings of depression according to yon Zerssen (DS/DS'), and mood ratings on a 100ram visual analogue scale (VAS) before and after light treatment (LT). Mean values ± SD axe indicated. Two DS/DS' ratings are given: a single value on the evening prior to the CR protocol, and the mean value of all 9 ratings during each CR. Mood ratings refer to the weighted 24-h mean of half-hourly ratings obtained throughout the 40-h CR (50mm = average mood).

with the corresponding surges in the 4-s high frequency activity. Five consecutive 4-s EEG spectra were averaged to obtain 20-s EEG spectra. This averaging procedure discarded 4-s epochs contaminated with outstanding high frequency activity in the 22.25-25.0 Hz band.

Within subjects, the same EEG derivation was used throughout the four nights. Power spectra in nonREM sleep were calculated by averaging the power density values of 20-s epochs scored as stage 2, 3, or 4. Only one of the 76 recordings had less than 4 h of sleep accumu- lated. Therefore, the analysis of sleep parameters was performed for the total sleep episode as well as for the first 4 h of sleep. In a first step, all data were separately analyzed for SAD patients and control women by analysis of variance (ANOVA) with the two repeated factors sleep deprivation (before/after) and LT (before/after). The data of all subjects were then pooled and analyzed by an ANOVA that included the two repeated factors and one grouping factor (SAD/control). Only 10 SAD patients were used in the ANOVA on spectral values because spectral data were not available in a single night due to a recording failure. For the statistical evaluation, the abso-

lute values of spectral power density were used. For visualization of the changes in spectral power after sleep deprivation or after LT the geometric mean of relative values were plotted. Paired t tests on the log-transformed relative values were used to test for differences between corresponding spectral data.

Results

Clinical Responses

With the exception of one severely ill woman, SAD patients were mildly depressed and without extreme atyp- ical symptoms (Table 1). After midday LT, all SAD patients fulfilled dual criteria for improvement (> 50% reduction in the HAM-D ratings to a value below eight; Terman et al 1989). It should be noted that although SAD patients improved markedly on all scales, and were euthy- mic after LT, they did not attain the low values of the controls (Table 1). This indicates a trait of slight residual depressivity in our group of SAD patients. Control women did not show a notable change in mood ratings.

Sleep EEG in SAD: Effects of Light and TSD BIOL PSYCHIATRY 489 1996;40:485-496

The response to total sleep deprivation (TSD) was assessed by comparing VAS mood ratings in the morning. TSD during the CR had an antidepressant effect in five of all eleven SAD patients, and these responders relapsed after recovery sleep. On average, SAD patients had sig- nificantly increased mood in the late morning (10:30-11: 30h) and in the evening (19:30-20:30h) of the second day of the CR compared with the values at the same times of the first day (p < 0.05; paired t test). During the CR after LT, when SAD patient had clinically improved, TSD had no significant effect on mood although four patients still responded to TSD with improved mood. Only one of the eight control women showed a beneficial response to TSD.

factor subject group were found as follows. TSD increased sleep efficiency to a larger extent in controls and reduced their amount of wakefulness within the first 4 h of sleep more than in SAD patients (group X TSD; p < 0.05). The effect of TSD on stage 2 sleep also differed between the groups (group X TSD; p < 0.05). Sleep latency was shortened after LT to a larger extent in controls (group × LT; p < 0.05). The 3-way interaction was also significant in this sleep parameter showing that the more pronounced shortening of sleep latency by LT in the controls was only present for the baseline night (group X TSD X LT; p < 0.05).

Sleep Parameters Mean values of sleep stage parameters are shown in Table 2. The upper part shows values calculated for the entire sleep episode and the lower part depicts sleep parameters for the first 4 h of sleep. For both subject groups, ANOVA revealed significant effects of TSD on most sleep stage parameters and a significant influence of LT on some sleep stages. Regardless of subject group, TSD increased total sleep time, sleep efficiency, and slow wave sleep, and it reduced sleep latency, REM sleep latency, the amount of wakefulness, and stage 1. Total minutes of REM sleep were increased after TSD whereas the percentage of REM sleep and the amount of REM sleep during the first 4 h of sleep were not significantly altered. After LT, total sleep time and sleep efficiency were increased and sleep latency was decreased in both groups. An increase in the amount and percentage of REM sleep after LT was noted in both subject groups but this effect reached significance only in SAD patients. The increase of REM sleep occurred evenly across the night as demonstrated by an analysis of REM sleep accumulation (data not shown). Controls showed an increase in stage 2 after LT, whereas stage 2 percent and the amount of stage 2 during the first 4 sleep hours were not significantly affected. In SAD patients, stage 1 during the first 4 hours of sleep was slightly but significantly reduced after LT. Significant interactions between the factors LT and TSD were found in the control women for sleep latency, slow wave sleep and stage 2 in the first 4 h of sleep. In SAD patients, only the interaction for the latter parameter was significant. All these interactions indicate that the effect of TSD was less pronounced after LT.

The 3-factor ANOVA with the grouping factor revealed that SAD patients had better sleep than the controls. They showed a higher sleep efficiency, longer total sleep time, and more stage 2 sleep during the entire sleep episode. SAD patients also accumulated less wakefulness during the first 4 h of sleep. Significant interactions with the

Electroencephalographic Power Spectra

Figures 2 and 3 show relative EEG power spectra aver- aged over nonREM sleep within the first 4 h of sleep. Absolute values of power density did not differ between SAD patients and controls in any frequency bin (group effect: p > 0.1; for all bins). The changes of EEG power density after 40 h of TSD are plotted in Figure 2. The sleep deprivation effect prior to LT is presented in the upper panel and the effect of TSD obtained after LT is shown in the lower panel. The spectral changes were similar for both subject groups and before and after LT. In all cases, the recovery nights had increased EEG activity in the frequencies between 0.25-11.0 Hz with peak values in the low delta band. In the control group, the increase did not reach significance in the 8.25-10.0 Hz range. A slight attenuation of EEG activity was present in the spindle frequencies. The decrease reached significance in the 12.25-14.0 Hz range for the SAD patients. Although for some frequency bands the effect of TSD appears to be more pronounced in the SAD-patients, the 3-factor ANOVA revealed no significant differences between the two groups (group X TSD: p > 0.05; for all frequencies). The prominent increases of power density in the delta frequencies reached higher values prior to LT (upper panel) than following LT (lower panel), however, signif- icant interactions between LT and TSD were found only in three isolated frequency bins for the SAD patients (2.75- 3.0 Hz; 10.25-11.0 Hz; 16.25-18 Hz) and in three isolated bins for the controls (1.75-2.0 Hz; 6.25-7.0 Hz; 14.25- 15.0 Hz). With respect to this interaction, the two groups did not differ significantly in any frequency bin (group X TSD X LT: p > 0.05).

The effects of LT on the EEG power spectrum of nonREM sleep are shown in Figure 3. Power values after LT are plotted as a percentage of the corresponding values obtained before LT. The upper panel depicts the effect of LT on the baseline night (BL) and the lower panel shows the LT effect for the recovery nights (REC). The devia-

490 BIOL PSYCHIATRY D.P. B runn e r et al 1996;40:485-496

I¢d¢2. SAD patients (N=11) Control women (N=8)

Before LT After LT ANOVA Before LT After LT ANOVA

BL REC BL REC TSD LT INT BL REC BL REC TSD LT INT

Total sleep episode

Timeinbed 510.8 509.9 510.1 509.9 488.1 485.8 485.7 484.3 (TIB) ±10.4 ±9.7 +10.9 ±10.0 ±9.1 ±9.4 ±9.4 ±9.4

Total sleep time 427.3 476.9 444.5 485.5 xx xx 335.3 437.9 376.0 457.0 (TST) ±13.5 ±9.7 ¢11.4 ±10.4 ±34.0 +12.4 ±24.1 ±10.8

Sleep efficiency 83.7 93.5 87.5 95.2 xx xx 68.7 90.2 77.2 94.3 100%*TST/TIB ±2.1 ±0.9 ±2.6 +0.8 ¢6.7 ±2.5 =4.4 ±1.0

Sleep latency 21.5 9.5 16.7 5.8 xx x 47.2 8.4 22.4 6.3 to stage 2 ¢2.7 ±2.9 ±5.6 ±0.8 ±12.9 ±1.2 ±7.2 ±1.1

REM sleep latency 112.4 67.6 105.1 55.4 xx 140.0 81.3 104.5 85.8 ±13.8 ±10.7 ±15.3 ±2.8 ±30.4 ±10.9 ±14.3 ±9.5

Wakefulness 54.1 15.3 41.7 10.5 x 100.8 33.1 82.9 15.7 after sleep onset ¢ 1 2 . 1 +5.2 +14.5 ¢4.0 +97.9 +13.0 ±18.2 ±4.2

Stage 1 41,3 22.4 38,3 19.5 xx 38.2 22.6 40.9 17.5 ±.2.8 :u?..1 ±2..8 ±2.6 ±6.7 ±4.1 ±5.1 ±3.3

Stage2 250.5 225.0 246.7 228.4 177.6 190.5 180.9 210.0 ±11.7 ¢11.1 ±9.9 ±10.1 ±24.4 ±16.4 ±20.1 ¢15.4

SIow wave sleep 60.3 121.5 62.9 121.6 xx 81.0 140.9 73.3 136.4 ±7.3 ±6.2 ±8.6 ,-7.1 ±11.5 ±18.0 ±12.2 ±16.2

REM sleep 75.2 108.0 96.5 115.9 x x 59.5 83.9 81.0 93.0 (REMS) ¢7.0 =9.0 ±9.0 ±8.0 ±14.4 ¢13.5 ¢6.8 ¢10.7

Percent REMS 17.7 22.5 21.7 23.7 x 18.6 19.0 21.8 20.2 (% of TST) ±1.5 ±1.7 ±1.8 ±1.3 ±3.0 ¢2.9 ±1.7 ±2.3

XX X

XX X

X XX XX

XX

XX

XX

XX X

First 4 hours of sleep

Wakefulness 26.9 4.7 13.7 2.2 xx 84.3 6.2 54.9 5.2 x after sleep onset ±6.8 ±2.2 ±3.7 ±1.4 +27.2 ±3.9 ±21.6 ±3.0

Stage 1 20.8 7.1 17.4 5.5 xx x 25.5 9.2 23.1 4.4 xx ±2.3 ±1.2 ±1.6 ±1.4 ±5.1 +1.9 ±2.7 ±1.1

Stage 2 142.5 92.0 131.9 97.0 xx xx 122.0 9 0 . 1 111.4 91.9 x ±4.0 ±4.9 ¢3.4 ±4.4 ±9.6 ±9.3 ±9.3 ¢12.2

Slow wave sleep 49.3 105.4 52.9 100.9 xx 54.2 109.7 85.7 108.7 xx ±5.1 ±4.1 ±4.9 ±4.8 +9.3 ¢12.2 ±10.4 ±12.5

REM sleep 27.3 35.5 37.8 36.5 x 32.8 31.0 39.8 35.0 ±4.6 ±3.6 ±4.5 ±3.1 ±5.5 +7.4 +4.2 ±6.1

Mean values (-+ sem) of sleep stage parameters for 11 female SAD-patients (left) and for 8 control women (right). Parameters are calculated for the entire sleep episode (upper part) and for the ftrst 4 h of sleep (lower part). Values of the baseline night (BL) and recovery night (REC) are listed for both light treatment (LT) conditions. Unless otherwise indicated the values represent minutes. For both, SAD patients and controls the results of a two-way ANOVA with the factors total sleep deprivation (TSD) and LT are shown. Significant interactions (INT) between the two factors are listed in the third column of the ANOVA. Significance levels of p < 0.05 (x) and p < 0.01 (xx) are indicated. Due to the skewed distribution of sleep latency values, the statistical evaluation of this parameter is based on the log-transformed values.

Sleep EEG in SAD: Effects of Light and TSD BIOL PSYCHIATRY 491 1996 ;40 :485 496

EFFECT of TOTAL SLEEP DEPRIVATION

180 -

. d

m 160

.E ~} 140

O

Q" 1 2 0 - "5

~ 100 ~. -

80-

60-

" 1 8 0 -

m ~ 160-

._~

Q. 1 2 0

"6

100 Q. .

8 0

6O

~ before LT

i , , , . '%-- - \ - sA0 " " . . . ~ . . . . . . . . CON

; ". . ~ : - , \ \',, .

o - - SAD - - ~ CON

f rSD*LT )

. . . . I ' ' ' ' I . . . . I . . . .

after LT

SAD

- - CON

( ] 'SO)

' ' I ' ' ' ' I . . . . I ' ' ' '

5 10 15 20 Frequency (Hz)

Figure 2. Effect of total sleep deprivation (TSD) before and after light treatment (LT) on EEG power spectra. Power density was calculated for nonREM sleep during the first 4 h of sleep. For each subject and frequency bin the values of the recovery nights were expressed relative to the corresponding values of the preceding baseline night. The geometric mean of these relative values are plotted for 11 SAD patients (SAD) and for eight control women (CON). Due to data loss in one night only 10 SAD patients contributed to the solid curve in the upper panel. The results of an ANOVA are indicated for each subject group separately. The frequency ranges with a significant effect of TSD are delineated by solid lines above the lower abscissa. Frequen- cies showing a significant interaction between TSD and LT are indicated above the upper abscissa (p < 0.05).

EFFECT of LIGHT TREATMENT (LT)

J=

- J I:o "6 "E o #_

130 -

120 -

110

100

90

8 0

BL nights

SAD

" J ~ . . . . . . . . CON

SAD (Hes t )

' ' ' • I ' ' ' ' I ' ' ' ' I ' ' ' '

O

J3

O ILl r r

"5 "E ,o

0 -

120 -

110 -

100

9 0 -

REC nights

" x /

. SAD (L'r) INT (GROUP*L ] )

80- ' ' ' ' I ' ' ' ' I ' ' ' ' I ' ' ' '

0 5 10 15 2O

Frequency (Hz)

Figure 3. Effect of light treatment (LT) on EEG power spectra in nonREM sleep for 11 SAD patients and eight control women (CON). In the upper panel power density values of the baseline night (BL) after LT are expressed relative to the corresponding values obtained in BL before LT. In the lower panel the values of the recovery night (REC) after LT are expressed relative to the values obtained in REC before LT. For each frequency bin and both subject groups the geometric mean of the relative values is plotted. Due to a loss of data in one recovery night only 10 SAD patients contributed to the solid curve in the lower panel. Statistical results are indicated above the abscissae by black bars (p < 0.05). The effect of LT was significant only for a few delta and theta frequencies in SAD patients and paired t-tests revealed significant differences only between the two BL nights. Three factor ANOVA revealed a significant interaction (INT) demon- strating that the two subject groups differed significantly in their response to LT in the delta and theta range (group*LT).

tions from the reference values were small and surpassed 10 percent only in the comparison of the two baseline nights in SAD patients. In control women, the 2-factor

ANOVA revealed no significant effect of LT across the entire frequency spectrum (LT: p > 0.1; for all bins). The

same analysis in SAD patients showed a significant effect of LT on EEG activity in the 1.25-1.5 Hz and 4.75-7.0 Hz

bands. As evaluated by paired comparisons, SAD patients had significantly increased EEG activity in the 0.75-6.0 Hz range for the baseline night whereas the changes in the recovery night did not reach significance in any frequency

bin. SAD patients and control women differed signifi- cantly in their response to LT in the 1.75-4.0 Hz and

4.75-6.0 Hz band (group X LT: p < 0.05).

492 BIOL PSYCHIATRY D.P. Brunner et al 1996;40:485-496

D i s c u s s i o n

The eleven SAD patients, who exhibited mild depressive symptoms, all improved after 5 days of midday light treatment. Since the study was not placebo controlled, but designed to analyze circadian and physiological effects of LT, the clinical effects can only be judged within the framework of the large number of studies showing an antidepressant effect of light in SAD (Terman et al 1989).

During the course of the CR protocol prior to LT, five of the 11 SAD patients responded to the sleep deprivation of this protocol and relapsed after recovery sleep. On average, SAD patients showed better mood after sleep deprivation, which contrasts with the slight decline in mood in the control group (see Wirz-Justice 1994b; Fig. 3). SAD patients have been shown to profit from sleep deprivation in other clinical trials (Wehr et al 1988). Thus, sleep deprivation reveals an important link between sleep regulation and mood in SAD as well as other forms of depression (Kuhs and Toelle 1991).

Sleep Stage Parameters

In the present study, SAD patients slept longer and better than control women, whereas REM sleep and slow wave sleep parameters did not differ between the two groups. Although control women did not report sleep disturbances, their sleep efficiency in the baseline nights was found to be rather poor. In comparison with published normative sleep data from healthy women of similar age (Williams et al 1974) sleep efficiency in our sample appears to be exceptionally low. Therefore, the findings of prolonged and better sleep in SAD patients might not hold true in a larger comparison with normals. It should be noted that our controls were carefully matched to cover the same wide age range as the SAD patients.

The first of two earlier studies that recorded sleep of SAD patients and control subjects in winter did not find significant differences in sleep continuity measures be- tween the two groups (Anderson et al 1994), but only a significant difference in slow wave sleep (SWS). This difference was not present in the comparison of all 26 subjects, but in only a subgroup of nine matched subjects. It is possible that a difference in mean age might have contributed to the larger amount of SWS in the controls. The second study reported differences between SAD patients and unmatched controls for several sleep param- eters (Palchikov et al 1996). A subgroup of 13 SAD patients had lower SWS during their winter depression, compared with seven controls. In both seasons, winter and summer, SAD patients had longer total sleep time and larger amounts of REM sleep than controls. These latter

differences indicate a trait marker for this group of SAD patients or, alternatively, selective recruitment of long sleepers into the patient group.

Taken together, there is no strong evidence that poly- somnographic sleep of SAD patients differs from that of healthy controls. In agreement with this notion, none of the sleep EEG studies in SAD (Rosenthal et al 1989; Endo 1993; Partonen et al 1993; Anderson et al 1994; Kohsaka et al 1994; Palchikov et al 1996) has found the pattern of sleep EEG changes that characterizes melancholic depres- sion. This notion is further supported by a direct compar- ison of polysomnographic sleep in seasonal and nonsea- sonal depressed patients (Thase 1989). Sleep in SAD patients was found to be similar to sleep of patients with anergic depression but showed little resemblance to sleep in endogenous depression.

Effects of Total Sleep Deprivation on Sleep

The analysis of the sleep stages revealed that the effects of TSD and LT on sleep were very similar for SAD patients and control women. Both subject groups exhibited the well-known regulatory response to one night of TSD as previously reported in healthy young men (Berger et al 1971; Borbtly et al 1981; Dijk and Czeisler 1993) and older subjects (Reynolds et al 1986). The increase in sleep efficiency and the reduction of wakefulness were signifi 7 cantly larger in the controls. Both of these interactions (group X TSD) are likely to be related to the lower level of sleep efficiency in the control women whose initially poorer sleep obviously can improve to a larger extent. The pronounced reduction in REM sleep latency after TSD in both of our groups is not present in the first recovery night in healthy young men (Berger et al 1971; Borbtly et al 1981; Dijk and Czeisler 1993) and might be specific for gender or age. The latter is supported by the reduced REM sleep latencies found after sleep deprivation in elderly men and women (Reynolds et al 1986; Bonnet 1986). The shortening of REM sleep latency could have been favored by the relatively long latencies in the baseline nights. Since a delayed onset of REM sleep is often seen in disrupted sleep, TSD may reduce REM sleep latency by consolidating sleep in the middle-aged and elderly. Con- versely, since an afternoon nap shortens REM sleep latency of the following sleep episode (Feinberg et al 1992; Werth et al 1994) one could also expect REM sleep latencies to be prolonged after TSD. This difference in the reactivity of Rem sleep latency between younger and older subjects, which might also reflect an increased sensitivity of the first nonREM sleep episode in the elderly to deficits in REM sleep, needs further investigation.

Sleep EEG in SAD: Effects of Light and TSD BIOL PSYCHIATRY 493 1996;40:485-496

Effects of Light Treatment on Sleep

After LT, significant improvement of sleep was observed in both subject groups. Apart from the reduction of sleep latency in the baseline night being more pronounced in control women, there was no difference in the response to LT between the two groups. Since in the present design the LT condition always followed upon the condition without treatment, an order effect is inevitably confounded with the effects of LT. In view of the demanding procedure of the CR protocol, an adaptation from the first to the second trial must be considered. Nevertheless, increased sleep efficiencies were also reported in a study in which the order of the active LT was counterbalanced (Anderson et al 1994). This finding was interpreted as a normalization of a decreased sleep efficiency in depressed SAD patients, although sleep efficiency levels in winter were not signif- icantly decreased compared to those in summer or in control subjects.

In the present investigation using midday LT, we did not find an increase in SWS which has been found after combined morning and evening LT (Anderson et al 1994). Using morning LT alone, an increase of SWS in the first 3 h of the night has been reported in one study (Endo 1993) whereas in another no modifications of sleep stage parameters were found (Partonen et al 1993). Common to all LT studies in SAD is the improvement of depressive symptoms whereas no large or consistent changes in sleep parameters emerge from available EEG data. The most frequently reported finding after LT in SAD is increased sleep efficiency. This effect is rather unspecific and could also result from the improvement of depressive symptoms and from mood-related changes in sleep and wake behav- ior. The present investigation revealed the same sleep stage changes in both groups, indicating that the consoli- dation of sleep was not necessarily related to the allevia- tion of depressive symptoms. The improvement of sleep was therefore induced either by the exposure to bright light and/or by the order of the recordings.

Increased sleep maintenance has been reported after evening LT in healthy elderly subjects (Campbell et al 1993). This timing of LT induced a phase delay of the circadian system which in itself is associated with an improvement in morning sleep maintenance. Only one other LT study with healthy subjects administered bright light in the afternoon when no phase shifts of circadian rhythms are expected (Carrier and Dumont 1995). Despite the large 'dose' of light administered in that experiment, no changes in sleep stage parameters were found. It was concluded that LT does not affect sleep when no phase shift occurs, or that good sleep in young subjects cannot be further improved by means of LT. In summary, none of the LT studies that reported changes in sleep parameters

could demonstrate that LT per se, independent of circa- dian, clinical, behavioral, and/or methodological contribu- tions alters sleep stages.

Effects of Total Sleep Deprivation on the Electroencephalographic Power Spectrum

In both groups of women power spectra of nonREM sleep were changed by TSD in a manner nearly identical to that previously found in healthy young men (Borbtly et al 1981; Dijk et al 1990, 1993). Increased EEG activity in all frequencies below 11 Hz was accompanied by slightly attenuated values in the spindle frequencies (12.25-14.0 Hz). As illustrated in Figure 2, this pattern of spectral changes was rather stable before and after LT. The two subject groups did not significantly differ in any frequency bin with respect to both the absolute power and the change induced by TSD. These data reveal that all EEG frequen- cies up to 20 Hz undergo predictable changes after TSD. In other words, the spectral results demonstrate that the homeostatic regulation of the nonREM sleep EEG is not altered in depressed SAD patients when compared to that in the remitted state or in control subjects. The increase of power density in the low delta frequencies is comparable or even larger than that reported after TSD in healthy young men (Borb61y et al 1981, Dijk et al 1993). This increase of power density in the delta frequencies demon- strates that the homeostatic build-up of nonREM sleep intensity is not reduced in middle-aged women. In agree- ment with another CR experiment (Dijk et al 1993) our data show that the effect of sleep deprivation on the EEG is not dependent on the level of physical activity.

Effects of Light Treatment on Electroencephalographic Power Spectrum

SAD patients showed significant changes in their non- REM sleep power spectrum after LT (Fig. 3). These changes were rather small and consisted of an increased power density in some delta and them frequencies. This effect was specific to SAD patients and cannot be ex- plained by the increased sleep efficiency after LT because sleep was equally improved in control subjects for whom no change in the EEG spectrum was detected. Significant effects of LT on the EEG of SAD patients were restricted to the baseline night but power density in the delta and theta frequencies tended to be increased also in the recovery night 56 hours after the last exposure to bright light. These lasting spectral changes, only seen in SAD patients, seem to be related to the lasting improvement of depressive symptoms, however they were very small

494 BIOL PSYCHIATRY D.P. Brunner et al 1996;40:485-496

compared with the spectral changes after TSD and are unlikely to account for the antidepressant effect. One possible explanation for the EEG changes could lie in the clinical remission during LT, which is characterized by increased motivation and energy to resume previously neglected activities. A curtailment of sleep, possibly re- lated to activation and rebound activities, was documented by the sleep diaries in the group of SAD patients. They reported a reduction of their daily sleep duration by an average of 73 minutes (_ 20; sem) on the 4 days prior to the second CR as compared with sleep duration on the 4 days prior to the first CR (p < 0.01; paired t test). In contrast, the controls reported an insignificant increase in their average daily sleep duration (12 _+ 29 minutes) during LT. Thus, only SAD patients carried a slight "sleep debt" into the second CR. In agreement with these sleep log data, the spectral changes after LT, which were observed only in SAD patients, are very similar in pattern and time course to the EEG changes found during recov- ery from repeated partial sleep deprivation (Brunner et al 1990, 1993a). This implies, that homeostatic mechanisms regulating the nonREM sleep EEG are sufficient to ex- plain the small changes in the EEG power spectra of our SAD patients. In the control women, as well as in healthy young men (Dijk et al 1989), repeated exposure to bright light had no significant effect on the EEG power spectrum. These findings suggest that LT has no impact on the sleep EEG of healthy subjects and, if at all, only marginal effects on the EEG of winter depressed patients.

Conclusions

In our comparative analyses of polysomnographic and spectral sleep EEG data of SAD patients and control subjects we could not find clear evidence of altered sleep or sleep regulation in winter depressed women. The SAD-specific small response of the sleep EEG to LT can be explained on the basis of normal homeostatic compen- satory responses to changes in sleep duration. It might be argued that the mild severity of depression in our SAD patients may account for the lack of changes in sleep and spectral EEG parameters, however, the present study agrees with previous reports demonstrating clinical im- provement without concomitant consistent changes in sleep parameters. Therefore, our EEG data do not support the assumption of an involvement of sleep mechanisms in the pathogenesis of winter depression. This conclusion is corroborated by earlier reports about sleep duration in SAD based on prospective sleep diary information which is more reliable than global retrospective reports (Shapiro et al 1994). Sleep diary data show that sleep duration is generally not altered after successful LT (Wirz-Justice et

al 1989; Shapiro et al 1994). When measured on a daily basis, sleep duration of SAD patients in Switzerland does not differ significantly between summer and winter (Brun- ner et al 1993b; Kr~iuchi et al 1993). Furthermore, daily estimates of sleep duration and timing do not correlate with therapeutic response (Wirz-Justice and Anderson 1990) and have no predictive value for the success of LT (Kr~iuchi et al 1993). Thus, most polysomnographic sleep data as well as some recent data on prospective daily ratings of sleep duration question the validity of hyper- sonmia as a necessary feature of SAD, unless patients are specifically selected for this characteristic.

In an analysis involving more than 100 SAD patients, multivariate regression analysis of seven individual SAD features revealed that only the level of social activities was significantly related to depression severity (Shapiro et al 1994). Similarly, Meesters et al (1993) reported a remark- ably parallel time course of depressive symptoms and of fatigue across LT and withdrawal. In view of these findings, changes in daytime energy and activity levels emerge as a hallmark in the clinical course of SAD patients, whereas changes in sleep parameters, including sleep duration, are difficult to demonstrate. The multiple sleep latency test in SAD has also failed to show evidence for increased sleep propensity during the day (Putilov et al 1995). Despite unchanged sleep propensity, sleep satiation due to naps and unintentional sleep episodes is likely to accompany winter depression because the inactivity and anergia of SAD patients strongly impair the ability to stay out of bed and to maintain wakefulness throughout the day. Mood-related anergia, inactivity, and hypersomno- lence result in behavioral sleep patterns that have second- ary effects on sleep parameters, however, if SAD patients are taken into the laboratory and are required to adhere to a predetermined sleep-wake schedule with strict avoidance of naps, sleep is found to be normal and not linked with the mood symptoms of SAD. This conclusion is supported by our EEG findings after sleep deprivation which demon- strate that sleep mechanisms are unchanged in SAD patients and are not involved in the antidepressant mech- anism of LT.

Remission from winter depression may be mediated by an ability of LT to increase activity and energy levels of SAD patients. Our finding that only SAD patients cur- tailed their sleep during LT at home also points to increased levels of energy and social activity in the course of clinical remission. Sleep curtailment during LT as documented by sleep diaries contradicts the finding of increased sleep time in the laboratory after LT. This apparent paradox is resolved by recognizing that SAD patients voluntarily shortened their sleep at home whereas in the laboratory their bedtimes were the same before and after LT. Thus, after remission from depression, sleep in

Sleep EEG in SAD: Effects of Light and TSD B1OL PSYCHIATRY 495 1996;40:485-496

the laboratory served as an opportunity to recover from previous sleep restriction similar to the compensatory sleep during the weekend in normal subjects.

Sleep changes across the clinical course in SAD pa- tients, like sleep changes across the days of the week in healthy subjects, can be explained by systematic variations in the sleep-wake pattern. These patterns are influenced by a person 's motivation, mood, and energy to be active and awake. LT is able to compensate for the anergia and depression in SAD-patients by a yet unknown mechanism. Our study, in which LT was given at midday, suggests that the antidepressant effect of LT does not involve a change in the phase of the circadian system nor a modification of the homeostatic regulation of nonREM sleep. The extent

to which the energizing and activating effects of LT are mediated by bright light itself, by a placebo mechanism, or by the enrollment in a study with intensive interpersonal interaction needs to be determined in future studies.

Support was received from the Swiss National Science Foundation grants #32-28741.90 and 32-32300.91 and the Roche Research Foundation.

The authors thank Carole Hetsch, Gabi Moll, Ellen Weber, Marielle Koenig, Vladimir Djurdjevic, and the student technicians for their help with data acquisition; Esther Werth, Dr. Christian Cajochen, and Dr. Peter Graw for their assistance in data analysis; Dr. Janis Anderson for initiating the study; Drs. Edith Holsboer-Trachsler and Alexander Bor- b~ly for generously allowing us to access their facilities; and the subjects for volunteering in this demanding study.

References Aitken RCB (1969): Measurement of feelings using visual

analogue scales. Proc Royal Soc Med 62:17-21.

Anderson JL, Rosen LN, Mendelson WB, Jacobsen FM, Skwerer RG, Joseph-Vanderpool JR, Duncan CC, Wehr TA, Rosenthal NE (1994): Sleep in fall/winter seasonal affective disorder: effects of light and changing seasons. J Psychosom Res 38:323-337.

Benca RM, Obermeyer WH, Thisted RA, Gillin JC (1992): Sleep and psychiatric disorders: a meta-analysis. Arch Gen Psychi- atry 49:651-668.

Berger RJ, Walker JM, Scott TD, Magnuson LJ, Pollack SL (1971): Diurnal and nocturnal sleep stage patterns following sleep deprivation. Psychonomic Science 23:273-275.

Bonnet MH (1986): Effect of 64 hours of sleep deprivation upon sleep in geriatric normals and insomniacs. Neuro~iol Aging 7:89-96.

Borb61y AA, Baumann F, Brandeis D, Strauch I, Lehmann D (1981): Sleep deprivation: Effect on sleep stages and EEG power density in man. Electroenceph Clin Neurophysiol 51:483-493.

Borb61y AA, Wirz-Justice A (1982): Sleep, sleep deprivation, and depression. Human Neurobiol 1:205-210.

Brunner DP, Dijk DJ, Tobler I, Borb61y AA (1990): Effect of partial sleep deprivation on sleep stages and EEG power spectra: Evidence for nonREM and REM sleep homeostasis. Electroenceph Clin Neurophysiol 75:492-499.

Brunner DP, Dijk DJ, Borb61y AA (1993a): Repeated partial sleep deprivation progressively changes the EEG during sleep and wakefulness. Sleep 16:100-113.

Brunner DP, Kr~iuchi K, Leonhardt G, Graw P, Wirz-Justice A (1993b): Sleep parameters in SAD: Effects of midday light, season, and sleep deprivation. Sleep lies 22:396.

Campbell SS, Dawson D, Anderson MW (1993): Alleviation of sleep maintenance insomnia with timed exposure to bright light. J Am Geriatr Soc 41:829-836.

Carrier J, Dumont M, (1995): Sleep propensity and sleep architecture after bright light exposure at three different times of day. J. Sleep Res 4:202-211.

Czeisler CA, Weitzman ED, Moore-Ede MC, Zimmerman JC,

Knauer RS (1980): Human sleep: Its duration and organiza- tion depend on its circadian phase. Science 210:1264-1267.

Czeisler CA, Kronauer RE, Mooney J J, Anderson JL, Allan JS (1987): Biologic rhythm disorders, depression, and photother- apy: A new hypothesis. Psy Clin NAmer 10:687-709.

Czeisler CA, Kronauer RE, Allan JS, Duffy JF, Jewett ME, Brown EN, Ronda JM (1989): Bright light induction of strong (type 0) resetting of the human circadian pacemaker. Science 244:1328 - 1333.

Dijk DJ, Czeisler CA (1993): Body temperature is elevated during the rebound of slow-wave sleep following 40-h of sleep deprivation on a constant routine. J Sleep Res 2:117- 120.

Dijk DJ, Beersma DGM, Daan S (1989): Bright morning light advances the human circadian system without affecting NREM sleep homeostasis. Am J Physiol 256:R106-111.

Dijk DJ, Brunner DP, BorbEly AA (1990): Time course of EEG power density during long sleep in humans. Am J Physiol 258:R650-R661.

Dijk DJ, Hayes B, Czeisler CA (1993): Dynamics of electroen- cephalographic sleep spindles and slow wave activity in men: effect of sleep deprivation. Brain Res 626:190-t99.

Eastman CI, Young MA, Fogg LF (1993): A comparison of two different placebo-controlled SAD light treatment studies. In Wetterberg L (ed), Light and Biological Rhythms in Man. Oxford: Pergamon Press, pp 371-383.

Endo T (1993): Morning bright light effects on circadian rhythms and sleep structure of SAD. Jikeikai Med J 40:295-307.

Feinberg I, Maloney T, March JD (1992): Precise conservation of NREM period 1 (NREMP1) delta across naps and noctur- nal sleep: Implications for REM latency and NREM/REM alternation. Sleep 15:400-403.

Gillin JC, Duncan WC, Pettigrew KD, Frankel BL, Snyder F (1979): Successful separation of depressed, normal, and insomniac subjects by EEG sleep data. Arch Gen Psychiatry 36:85-90.

Hamilton M (1967): Development of a rating scale for primary depressive illness. Br J Soc Clin Psychol 6:278-296.

Haug HJ, Fiihndrich E (1986): Problems in defining response in therapy studies. Pharmacopsychiat 19:170-171.

496 BIOL PSYCHIATRY D.P. Brunner et al 1996;40:485-496

Haug HJ, Wirz-Justice A, Kr~iuchi K, Graw P, Hetsch C, Leonhardt G, Brunner DP (1996): Circadiane Veranderungen der Stimmung bei Depressiven und Gesunden wahrend einer "Constant-Routine"-Untersuchung. In M611er HJ, MiJller- Spahn F, Kurtz G (eds), Biologische Psychiatrie 1994. Wien: Springer Verlag, in press.

Kohsaka M, Honma H, Fukuda N, Kobayashi R, Honma K (1994): Does bright light change sleep structures in seasonal affective disorder? Soc Light Treatment Biol Rhythms (Abstr) 6:32.

Krauchi K, Wirz-Justice A, Graw P (1993): High intake of sweets late in the day predicts a rapid and persistent response to light therapy in winter depression. Psychiatry Res 46:107- 117.

Kuhs H, Toelle R (1991): Sleep deprivation therapy. Biol Psychiatry 29:1129-1148.

Lewy AJ, Sack R, Miller L, Hoban T (1987): Antidepressant and circadian phase-shifting effects of light. Science 235:352- 354.

Meesters Y, Jansen JHC, Lambers PA, Bouhuys AL, Beersma DGM, Van den Hoofdakker RH (1993): Morning and evening light treatment of seasonal affective disorder: Response, relapse and prediction. J Affect Disord 28:165-177.

Palchikov VE, Zolotarev DY, Danilenko KV, Putilov AA (1996): Effects of the seasons and of bright light administered at different times of day on sleep EEG and mood in patients with seasonal affective disorder. Biol Rhythm Res, in press.

Partonen T, Appelberg B, Partinen M (1993): Effects of light treatment on sleep structure in seasonal affective disorder. Eur Arch Psychiatry Clin Neurosci 242:310-313.

Putilov AA, Danilenko KV, Palchikov VE, Schergin SM (1995): The multiple sleep latency test in seasonal affective disorder: no evidence for increased sleep propensity. Soc Light Treat- ment Biol Rhythms (Abstr) 7:30.

Rechtschaffen A, Kales A (1968): A Manual of Standardized Terminology, Techniques, and Scoring System for Sleep Stages of Human Subjects. National Institutes of Health, Publication No. 204. Washington, DC: U.S. Government Printing Office.

Reynolds CF, Kupfer DJ (1987): Sleep research in affective illness: State of the art circa 1987. Sleep 10:199-215.

Reynolds CF, Kupfer D J, Hoch CC, Stack JA, Houck PR, Berman SR (1986): Sleep deprivation in healthy elderly men and women: Effects on mood and on sleep during recovery. Sleep 9:492-501.

Rosen LN, Targum SD, Terman M, Bryant MJ, Hoffman H, Kasper SF, Hamovit JR, Docherty JP, Welch B, Rosenthal NE (1990): Prevalence of seasonal affective disorder at four latitudes. Psychiatry Res 31:131-144.

Rosenthal NE, Sack DA, Gillin JC, Lewy AJ, Goodwin FK, Davenport Y, Mueller PS, Newsome DA, Wehr TA (1984): Seasonal Affective Disorder: A description of the syndrome and preliminary findings with light therapy. Arch Gen Psy- chiatry 41:72-80.

Rosenthal NE, Skwerer RG, Levendosky BA, Joseph-Vander- pool JR, Jacobsen FM, Duncan CC, Gaist PA, Wehr TA

(1989): Sleep architecture in seasonal affective disorder: The effects of light therapy and changing seasons. Sleep Res 18:440.

Shapiro CM, Devins GM, Feldman B, Levitt AJ (1994): Is Hypersomnolence a feature of seasonal affective disorder? J Psychosom Res 38 (Suppl 1):49-54.

Society of Light Treatment and Biological Rhythms (1990): Consensus statement on the efficacy of light treatment for SAD. LTBR 3:5-9.

Terman M (1993): Problems and prospects for use of bright light as a therapeutic intervention. In Wetterberg L (ed), Light and Biological Rhythms in Man. Oxford: Pergamon Press, pp 421-436.

Terman M, Terman JS (1991): Seasonal Affective Disorder and Light Therapy: Report to the Depression Guidelines Panel. P.H.S. Agency for Health Care Policy and Research.

Terman M, Terman JS, Quitkin FM, McGrath PJ, Stewart JW, Rafferty B (1989): Light therapy for seasonal affective disorder: A review of efficacy. Neuropsychopharmacol 2:1- 22.

Thase ME (1989): Comparison between seasonal affective dis- order and other forms of recurrent depression. In Rosenthal NE, Blehar MC (eds), Seasonal Affective Disorders and Phototherapy. New York: Guilford Press, pp 64-78,

Wehr TA, Rosenthal NE, Sack DA (1988): Environmental and behavioral influences on affective illness. Acta Psychiat Scand (Suppl 341) 77:44-52.

Werth E, Dijk DJ, Achermann P, Borb61y AA (1994): Effect of an early evening nap on sleep and the sleep EEG. J Sleep Res (Suppl 1) 3:276.

Williams RL, Karacan I, Hursch CJ (1974): Electroencephalog- raphy (EEG) of Human Sleep: Clinical Applications. New York: John Wiley & Sons.

Wirz-Justice A (1994a): A decade of light therapy for seasonal affective disorder. In Jung EG, Holick MF (eds), Biological Effects of Light. Berlin: Walter de Gruyter, pp 192-205.

Wirz-Justice A (1994b): Biological rhythms in mood disorders. In Bloom FE, Kupfer DJ (eds), Psychopharmacology. The Fourth Generation of Progress. New York: Raven Press, pp 999-1017.

Wirz-Justice A, Anderson JL (1990): Morning light exposure for the treatment of winter depression: the one true light therapy? Psychopharmacol Bull 26:511-520.

Wirz-Justice A, Graw P, Kr~iuchi K, P61dinger W (1989): Phototherapy in Switzerland: "Mehr Licht". In Thompson C, Silverstone T (eds), Seasonal Affective Disorder. London: CNS Publishers, pp 169-185.

Wirz-Justice A, Graw P, Kr~iuchi K, Gisin B, Jochum A, Arendt J, Fisch HU, Buddeberg C, P61dinger W (1993): Light therapy in seasonal affective disorder is independent of time of day or circadian phase. Arch Gen Psychiatry 50:929-937.

von Zerssen D, Koeller DM (1976): Paranoid-Depressivitii- tsskala. Beltz Testgesellschaft, Weinheim Germany.

Zulley J, Wever R, Aschoff J (1981): The dependence of onset and duration of sleep on the circadian rhythm of rectal temperature. Pliigers Arch 391:314-318.